What Qualitative Data Can Reveal About Inflation

A Hot Air Balloon Being Inflated

Takeaway

This study introduces a framework using LLMs to translate qualitative business commentaries from the Federal Reserve’s "Beige Book" into structured economic data. By feeding firm-level explanations into a macroeconomic model, the authors pinpoint the precise triggers driving historical inflation surges.

---

Key Points

• Listening over estimating: Qualitative text allows economists to observe why firms say they changed prices, capturing conditional triggers that standard numerical data misses.

• Input costs led the charge: Non-labor input costs (such as raw materials and energy) accounted for over 40% of post-2019 inflation variation, far outstripping the direct impact of consumer demand.

• A multi-stage inflation story: Post-pandemic inflation evolved in waves—shifting from an initial demand recovery to a prolonged push from input costs, followed by lagging wage pressures.

---

When prices at the grocery store or gas pump suddenly surge, economists often debate the culprit. Was it government stimulus driving up consumer demand? Supply chain bottlenecks? Or rising labor costs?

Standard macroeconomic models try to infer these answers by running statistical tests on aggregate time-series numbers like CPI or GDP. But numbers alone struggle to explain causality: if both sales and prices rise, did strong demand push prices up, or did rising costs force prices higher despite steady demand?

To solve this, researchers Chenyu Hou, Jiannan Jiang, and Tao Wang took a bottom-up approach: directly analyzing how businesses explain their own pricing decisions in real time.

Research

The authors set out to determine what economic forces actually trigger firms to adjust prices and how these micro-level explanations shape macro inflation dynamics.

They analyzed over 3,300 regional reports from the Federal Reserve’s "Beige Book"—a qualitative summary of economic conditions compiled eight times a year since 1970 based on interviews with business contacts across the United States. Using an advanced Large Language Model (GPT-5), the team extracted explicit causal statements (attributions) linking price decisions to underlying factors (e.g., "raised shipping prices due to high demand" or "held prices steady despite rising labor costs").

They then embedded these narrative statistics into a economic model featuring "menu costs"—the friction firms face when changing prices—to isolate aggregate inflation drivers from 1990 through 2024.

Findings

The researchers found that qualitative business narratives track real-world economic dynamics with remarkable accuracy. In particular, firm narratives closely mirror the frequency of price adjustments across the economy—capturing how often businesses alter their price tags during periods of high economic uncertainty.

When analyzing the post-2019 inflation surge, the model revealed a distinct multi-phase progression:

1. Early 2020: Inflation dropped sharply due to collapsed demand during COVID-19 lockdowns.

2. Late 2020 to Early 2021: A temporary rebound in demand briefly pushed prices upward.

3. March 2021 onward: Non-labor input costs (such as energy, freight, and raw materials) became the overwhelming driver of inflation, coinciding with core CPI breaching the Fed’s 2% target.

4. Late 2021 to 2022: Wage pressures emerged as a secondary, longer-lasting contributor to elevated inflation.

Overall, the model attributes more than 80% of post-2019 inflation variation to three factors: non-labor input costs accounted for over 40%, wages explained about 25%, and direct demand pressures contributed just 14%.

Why did demand play such a modest direct role in repricing? Because changing prices involves administrative effort and customer friction, firms rarely reprice just because demand shifts slightly. Instead, sharp cost increases act as the tipping point that forces businesses to update their price tags.

Limitations

The Beige Book relies on qualitative interviews rather than a statistically random sample of all U.S. businesses. Additionally, firms do not report every factor behind every decision, creating "reporting frictions" that required structural adjustments in the model. Finally, the framework assumes symmetric effects for price increases and decreases, though downward price cuts occur less frequently in practice.

Why It Matters

If policymakers misdiagnose the cause of inflation—for instance, aggressively cooling demand when price spikes are primarily driven by input cost shocks—they risk causing unnecessary economic downturns. By bridging qualitative text with formal macroeconomics, this study provides central bankers with a framework to harness real-time business narratives, leading to more accurate diagnoses of monetary policy challenges.

Learn More

• Paper Title: Inflation Drivers in Firms’ Words: Bridging Micro Narratives and Macro Dynamics

• Authors: Chenyu (Sev) Hou, Jiannan (Jay) Jiang, and Tao Wang

• Publication: Working Paper (Simon Fraser University, University of Texas at Austin, Bank of Canada)

• Year: 2026

• URL: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7134398

Econ Today Explains

Economic Concept: Menu Costs "Menu costs" refer to the literal and figurative costs a business incurs when changing its prices—named after a restaurant having to print new physical menus. These costs include updating software, re-tagging inventory, analyzing competitors, and managing customer dissatisfaction.

Because altering prices takes time and money, firms do not adjust prices continuously in response to small economic shifts. Instead, prices remain sticky until an economic shock (like a large spike in material costs) is big enough to outweigh the menu cost, triggering a sudden, discrete price change. Understanding menu costs helps economists analyze the "extensive margin" of inflation—how frequently price tags change across the economy.

This article was written by AI but reviewed by a real human.